Update Event protocol de/serialization to follow our specification

[kamilogorek]

Currently, there are multiple places where we diverge with our own spec, which makes our SDK non-compliant.
This is somewhat problematic, as the SDK is used internally by tools like sentry-cli, which should be able to send any arbitrary, but the still valid event that was somehow stored locally.

Currently reported issues:

• context inferred "type" field not being respected #483
• request.data field should support structured data #484
• request query string & cookies should support structured data #485
• timestamp field no respecting documentation #486
• logentry should support logentry object based on docs #487

@boxxxie thanks for the reports, I allowed myself to transfer all of them here.
I'll treat your opened issues as separate references, and this issue as aggregated summary. This way it'll be much easier for me to keep track of how many things are still there to be fixed.
If you find any more issues, please open separate issues in this repo, just as you did before, so I can add them to the list.

Spec reference: https://github.com/getsentry/sentry-rust/blob/master/sentry-types/src/protocol/v7.rs

[boxxxie]

i made all of those error reports, mainly because i found sentry-cli the best tool to use (i like dealing with json, rather than having a program/SDK build my data for me). i had a lot of issues with lack of documentation with sentry, until i found the SDK docs & official json schema.

after finding these initial errors, which were from real world test data from another sentry project, i wrote a data generator that makes valid payloads based on the json schema, and validated against the json schema.

my program currently isn't setup to produce minimal error cases, maybe this can be fixed without much work. would you guys be interested in a repo that details reports from this sorta fuzz testing i have created?

[kamilogorek]

Sure, any help is appreciated! :)

[boxxxie]

i have my data generation working, but i need to tweak it a bit or the size/time explodes, though maybe that's only a problem for my validation step. i guess i need to profile a bit. currently when i generate 100 examples it takes 16 minutes, even though 1 example takes a few milliseconds. this is my first time using this style of testing (test-check).

in the next few days i should have something that is fast at producing minimal error examples